Chang et al.

Vol 443|21 September 2006|doi:10.1038/nature05053
The cause of the fragile relationship between the
Pacific El Niño and the Atlantic Niño
Ping Chang1, Yue Fang1, R. Saravanan2, Link Ji1 & Howard Seidel1
El Niño, the most prominent climate fluctuation at seasonal-tointerannual timescales, has long been known to have a remote
impact on climate variability in the tropical Atlantic Ocean, but a
robust influence is found only in the northern tropical Atlantic
region1. Fluctuations in the equatorial Atlantic are dominated by
the Atlantic Niño2,3, a phenomenon analogous to El Niño, characterized by irregular episodes of anomalous warming during
the boreal summer. The Atlantic Niño strongly affects seasonal
climate prediction in African countries bordering the Gulf of
Guinea4,5. The relationship between El Niño and the Atlantic Niño
is ambiguous and inconsistent. Here we combine observational
and modelling analysis to show that the fragile relationship is a
result of destructive interference between atmospheric and oceanic
processes in response to El Niño. The net effect of El Niño on the
Atlantic Niño depends not only on the atmospheric response that
propagates the El Niño signal to the tropical Atlantic, but also on a
dynamic ocean–atmosphere interaction in the equatorial Atlantic
that works against the atmospheric response. These results
emphasize the importance of having an improved ocean-observing system in the tropical Atlantic, because our ability to predict
the Atlantic Niño will depend not only on our knowledge of
conditions in the tropical Pacific, but also on an accurate estimate
of the state of the upper ocean in the equatorial Atlantic.
There is clear observational evidence that the anomalous warming
in the eastern equatorial Pacific during an El Niño produces a
warming signal in the troposphere as a result of increased atmospheric heating6, which then propagates rapidly eastward in the form
of an equatorial Kelvin wave and westward in the form of a Rossby
wave, as predicted by simple theoretical models7,8 (Fig. 1a). The
tropospheric warming stabilizes the environment, causing reduced
moist convection, which acts in conditionally unstable environments
to redistribute energy from the boundary layer to the free troposphere. As a result, boundary layer energy, primarily in the form of
latent heat that originates via evaporation from the ocean surface,
will accumulate9. This then provides a ‘back pressure’ that reduces
evaporation from the ocean surface to the boundary layer, leading to
a warming of the ocean mixed layer. This so-called ‘tropospheric
temperature mechanism’ is arguably an important driving mechanism for the tropical El Niño teleconnection10,11. In the deep tropical
Atlantic, it works most effectively from late boreal winter to early
spring when the sea surface temperature (SST) is warmest seasonally
and the Intertropical Convergence Zone is closest to the Equator. The
overall reduction of precipitation in the tropical Atlantic basin
observed following El Niños is consistent with the weakening of
moist convection12. Modelling studies11,13 consistently show that in
the absence of ocean dynamics a basinwide warming of the tropical
Atlantic follows an El Niño.
However, the regressed SST response derived from observations
Figure 1 | Tropospheric and surface temperature response of the tropical
Atlantic to El Niño revealed by a regression analysis. a, The regressed
anomalous tropospheric temperature averaged between atmospheric
pressure levels of 800 mbar and 200 mbar over January–April against the
previous boreal winter, December–January, NINO3 SST (SST anomaly
averaged over 1508 W–908 W and 58 S–58 N). b, The similar regression of
the observed March–June SST anomaly. The SST response is expected to lag
the tropospheric temperature response because it is driven by surface heat
fluxes. The mid-tropospheric temperatures were derived from the 43-year
(1958–2000) reanalysis product, ERA40, of the European Centre for
Medium-Range Weather Forecasts (ECMWF);
research/era and the SST is from the Reynolds Optimum Interpolation SST22
during the same period. The coloured shading shows areas that exceed a 99%
significance level based on Student’s t-test.
Department of Oceanography, 2Department of Atmospheric Sciences, Texas A&M University, College Station, Texas 77843, USA.
© 2006 Nature Publishing Group
NATURE|Vol 443|21 September 2006
Figure 2 | Atmospheric response to the 1982/83 and 1997/98 El Niños. a–c, 1982/83; d–f, 1997/98. The left panels show the middle troposphere
temperature anomaly averaged between atmospheric pressure levels of 800 mbar and 200 mbar, based on the ECMWF ERA40 reanalysis product. The right
panels show the observed SST and surface wind stress (t x) anomaly from the same ECMWF and Reynolds SST products.
© 2006 Nature Publishing Group
NATURE|Vol 443|21 September 2006
does not show a basinwide warming (Fig. 1b), as one would expect
from the tropospheric temperature mechanism. The robust warming
is found only in the north tropical Atlantic. This may be attributed
partially to the fact that the tropospheric temperature mechanism
works more efficiently in this region because of the warmer mean SST
(ref. 11) and that El Niños can cause a weakening in the northeasterly
trade wind, resulting in a further reduction in evaporative heat loss
and thus enhancing the warming12. In the equatorial and south
tropical Atlantic regions, however, the observed response is more
perplexing: it tends to be opposite to that in the troposphere, albeit
less statistically significant.
The lack of a robust and consistent response of the equatorial SST
to El Niño can be made more evident by contrasting the tropospheric
and surface response to the 1982/83 and 1997/98 El Niños, two of
the strongest events in the instrumental record (Fig. 2). In spite of
the well-defined tropospheric warming in both events, the
surface responses are drastically different along the Equator and
along the southern African coast: persistent surface warming was
observed following the 1997/98 El Niño, consistent with the tropospheric temperature mechanism, while a cooling condition prevails
in the wake of the 1982/83 El Niño. We wondered what physical
factors cause the equatorial Atlantic to respond differently to these
El Niños.
First, we considered the surface wind response in the western
tropical Atlantic region. In the case of the 1982/83 El Niño, there is a
well-defined easterly wind anomaly that persists from the late boreal
winter to the early boreal summer in this region (Fig. 2). Such a wind
Figure 3 | Observed response of tropical Atlantic zonal wind stress to
El Niño. a, Correlation between the December–January NINO3 index and
the zonal wind stress anomaly from the ECMWF ERA-40 reanalysis product
over the tropical Atlantic sector in the following January–April from
1958–2000. The blue contours indicate negative and the red contours
positive correlation. The colour shade shows areas that exceed a 95%
significance level based on Student’s t-test. b, Normalized time series of the
January–February wind stress index (with reversed sign) over the western
equatorial region (608 W–208 W and 58 S–58 N) indicated by the rectangle in
a, and the December–January NINO3 index. The correlation coefficient
between the two indices is 0.6. The blue vertical bars indicate those El Niño
events (1965/66, 1976/77, 1982/83, 1986/87 and the 1991/92 El Niño) that
produced a strong easterly wind anomaly and the red indicate those
(1963/64, 1968/69, 1972/73, 1987/88, 1994/95 and the 1997/98 El Niños)
that did not.
response is apparently much weaker during the 1997/98 El Niño. The
relationship between the wind variability over the western tropical
Atlantic and El Niño has previously been noted14–16. A correlation
analysis between the observed NINO3 SST time series and the zonal
wind stress anomaly reveals an extended area along the western
equatorial Atlantic where the correlation coefficient is significantly
negative for the months following the peak of El Niño (Fig. 3a). A
closer examination reveals that all the El Niño events that did not
produce a warming in the equatorial Atlantic were accompanied by
anomalously strong easterly winds in this region, while those El Niños
that did produce a warming did not have such strong wind anomalies
(Fig. 3b).
A previous modelling study16 suggested that El Niño may lead to
cooling in the equatorial Atlantic. Dynamically, the equatorial cooling can be understood in terms of the Bjerknes feedback where, in
this case, an easterly wind stress anomaly along the Equator acts to
shoal the thermocline, enhancing the stratification and at the
same time increasing the vertical entrainment rate on the eastern
Figure 4 | The simulated tropical Atlantic response to El Niño in three
ensembles of numerical experiments. a, b, Ensemble averaged regressions
against the December–January NINO3 index from the POGA-CL runs for
the January–April troposphere temperature (8C) averaged between
atmospheric pressure levels of 800 mbar and 200 mbar (a) and surface winds
(vectors; b), where the blue contours show the negative correlation of the
zonal surface-wind with the NINO3 index and the red contours the positive.
(c–f), Regressions of the ensemble averaged SST in March–April and
May–June against the December–January NINO3 index from POGA-ML
and POGA-RG runs. The colour shade shows areas that exceed a 99%
significance level based on Student’s t-test.
© 2006 Nature Publishing Group
NATURE|Vol 443|21 September 2006
side of the basin, producing anomalous cooling at the surface. The
atmosphere responds to this cooling by further strengthening the
wind anomaly, forming a positive feedback between the ocean and
atmosphere. We hypothesize that the cooling produced by the
Bjerknes feedback competes with the tropospheric-temperatureinduced warming. This competition is the main cause of the fragile
relationship between the Pacific El Niño and the Atlantic Niño.
Borrowing an analogy from wave theory, we refer to this process as
‘destructive interference’.
It is difficult to test the destructive interference hypothesis solely
on the basis of observational analysis, because observations always
capture coupled variability and provide only a single realization of
stochastic climate phenomena. We therefore turn to ensembles of
coupled- and uncoupled-climate-model experiments (see Methods
for model descriptions). The first ensemble of experiments—hereafter referred to as POGA-CL (see Methods)—is designed to examine
the direct tropospheric and surface response of the atmosphere to El
Niño. Figure 4a and b shows the atmospheric model ensemble mean
response. It captures not only the observed structure of the tropospheric warming associated with El Niño (Fig. 1a), but also the
surface wind response (Fig. 3a). In particular, the easterly wind
anomaly in the western basin that is critically important to the
Bjerknes feedback is well-simulated, indicating that the direct
atmospheric response to El Niño is already sowing the seeds of the
destructive interference.
We then conducted two additional ensembles of coupled-model
experiments (hereafter referred to as POGA-ML and POGA-RG,
respectively; see Methods). These experiments are specifically
designed to test the destructive interference hypothesis by including
both the thermodynamic and dynamic ocean–atmosphere interactions in the simulations. POGA-ML experiments, in which only
the thermodynamic interaction is permitted, result in a surface
warming taking place shortly after the peak of El Niño. The warming
persists through the summer months along the Equator and along
the southern African coast, as anticipated by the tropospheric
temperature mechanism (Fig. 4c, d). Some of the warming along
and to the north of the Equator can also be attributed to windinduced latent heat flux changes12. In contrast, POGA-RG experiments, in which the dynamic interaction is also permitted, reveal a
much weaker boreal spring equatorial warming and a cooling
tendency during the early boreal summer (Fig. 4e, f). This cooling
effect is attributed to the dynamic ocean response to the easterly
surface wind anomaly associated with the warming in the troposphere. Further analyses indicate that the vertical advection of heat is
primarily responsible for the cooling. A recent comprehensive
coupled-climate study reports a similar finding17. Therefore, these
modelling results support the destructive interference hypothesis as a
plausible cause of the fragile relationship between the Pacific El Niño
and the Atlantic Niño.
It is unclear what causes the winds in the western equatorial
Atlantic to respond strongly to some El Niños, but not others, in spite
of a significant overall correlation. One factor may be related to the
pre-existing tropical Atlantic SST condition prior to an El Niño18.
Model experiments where the annual cycle of Atlantic SST in the
POGA-CL runs was replaced by the observed Atlantic SSTshow that a
pre-existing warm SST anomaly may substantially weaken the wind
response, resulting in a noticeable drop in the overall correlation
between the western Atlantic wind index and NINO3. Another factor
is related to stochasticity of the atmosphere. We noted that the
correlation between the wind index derived from an individual
ensemble member of the POGA-CL runs and the NINO3 can
decrease from the ensemble mean correlation of over 0.8 to about
0.6 for individual members. Additionally, the wind response can
depend on the atmospheric heating structure and duration, which
can vary considerably from one El Niño to another.
The Atlantic Niño is not a simple passive response to the Pacific El
Niño, but a complex one involving destructive interference between
Pacific remote influence and Atlantic ocean–atmosphere feedback,
making it a challenge for seasonal climate prediction. The SST
anomaly in the equatorial Atlantic—a key climate variable for
seasonal climate forecasts—depends on the relative dominance of
the tropospheric-temperature-induced heating compared to the
dynamic ocean–atmosphere feedback. Therefore, a successful forecast of SST not only requires that climate models simulate both
processes accurately, but also requires an accurate estimate of the
atmospheric state, both in terms of temperature and surface wind
response to El Niño, and of the upper ocean state. The latter points to
the necessity of an ocean observing system in the tropical Atlantic
because an accurate estimate of the upper ocean state is crucial for
determining the strength of the dynamic ocean–atmosphere feedback. In contrast, the ocean dynamics appear less critical in the north
tropical Atlantic, for which even an atmospheric general circulation
model coupled to a simple slab ocean is useful in forecasting seasonal
SST anomalies19.
Models. The atmospheric model is the Community Climate Model version 3
(CCM3) developed at the National Center for Atmospheric Research with the
standard configuration of T42 spectral truncation in the horizontal and 18
hybrid levels in the vertical. It incorporates a comprehensive suite of physical
parameterizations including a non-local boundary layer parameterization and
improved radiative and convection parameterizations20. The coupled model
is comprised of the CCM3 as the atmospheric component and an extended
1.5-layer reduced-gravity ocean model that has been used extensively in the
study of the Pacific El Niño21 and also applied to the study of the Atlantic Niño3.
The ocean model has a resolution of 28 in longitude by 18 in latitude and is
coupled to the CCM3 in the tropical Atlantic sector (between 308 S and 308 N),
outside which the atmospheric model is forced with observed SSTs.
POGA-CL experiments. The Pacific Ocean-Global Atmosphere (POGA)Climatological (CL) ensemble simulation consists of nine runs where the CCM3
was forced with observed SSTs from 1950 to 1995 in the Pacific and with the
annual cycle of SST in the Atlantic. Each ensemble member differs only slightly
in its atmospheric initial condition.
POGA-ML experiments. The POGA-Mix Layer (ML) ensemble simulation
consists of 12 runs where, in the Pacific sector, the coupled model was forced
with observed SSTs from 1980 to 2000. In the tropical Atlantic sector where the
model is coupled, the ocean component had all the dynamic processes (except
diffusion) disabled, so that the ocean model essentially degenerates to a slab
ocean where SST changes are completely determined by atmospheric surface
heat fluxes. As in POGA-CL, each ensemble member differs only slightly in its
atmospheric initial condition.
POGA-RG experiments. The POGA-Reduced Gravity (RG) ensemble simulation is identical in set-up to POGA-ML, except that all the dynamic processes
of the reduced-gravity ocean model are retained.
Received 3 November 2005; accepted 6 July 2006.
Enfield, D. B. & Mayer, D. A. Tropical Atlantic SST variability and its relation to
El Niño–-Southern Oscillation. J. Geophys. Res. 102, 929–-945 (1997).
2. Merle, J. Variabilité thermique annuelle et interannuelle de l’océan Atlantique
équatorial Est. L’hypothése d’un “El Niño” Atlantique. Oceanol. Acta 3,
209–-220 (1980).
3. Zebiak, S. E. Air–-sea interaction in the equatorial Atlantic region. J. Clim. 6,
1567–-1586 (1993).
4. Goddard, L. & Mason, S. J. Sensitivity of seasonal climate forecasts to persisted
SST anomalies. Clim. Dyn. 19, 619–-632 (2002).
5. Giannini, A., Saravanan, R. & Chang, P. Oceanic forcing of Sahel rainfall on
interannual to interdecadal time scales. Science 302, 1027–-1030 (2003).
6. Yulaeva, E. & Wallace, J. M. Sensitivity of seasonal climate forecasts to
persisted SST anomalies. J. Clim. 7, 1719–-1736 (1994).
7. Gill, A. E. Some simple solutions for heat-induced tropical circulation. Q. J. R.
Meteorol. Soc. 106, 447–-462 (1980).
8. Saravanan, R. & Chang, P. Interaction between tropical Atlantic variability and
El Niño–-Southern Oscillation. J. Clim. 13, 2177–-2194 (2000).
9. Brown, R. G. & Bretherton, C. S. A test of the strict quasi-equilibrium theory on
long time and space scales. J. Atmos. Sci. 54, 624–-638 (1997).
10. Chiang, J. C. H. & Sobel, A. H. Tropical tropospheric temperature variations
caused by ENSO and their influence on the remote tropical climate. J. Clim. 15,
2616–-2631 (2002).
11. Chiang, J. C. H. & Lintner, B. R. Mechanisms of remote tropical surface
warming during El Niño. J. Clim. 18, 4130–-4149 (2005).
© 2006 Nature Publishing Group
NATURE|Vol 443|21 September 2006
12. Giannini, A., Chiang, J. C. H., Cane, M. A., Kushnir, Y. & Seager, R. The ENSO
teleconnection to the tropical Atlantic Ocean: contributions of the remote and
local SSTs to rainfall variability in the tropical Americas. J. Clim. 15,
4530–-4544 (2001).
13. Alexander, M. & Scott, J. The influence of ENSO on air–-sea interaction in the
Atlantic. Geophys. Res. Lett. 29, doi:10.1029/2001GL014347 (2002).
14. Delecluse, P., Servain, J., Levy, C., Arpe, K. & Bengtsson, L. On the connection
between the 1984 Atlantic warm event and the 1982–-1983 ENSO. Tellus A 46,
448–-464 (1994).
15. Carton, J. A. & Huang, B. Warm events in the tropical Atlantic. J. Phys.
Oceanogr. 24, 888–-903 (1994).
16. Latif, M. & Barnett, T. P. Interactions of the tropical oceans. J. Clim. 8,
952–-964 (1995).
17. Huang, B. Remotely forced variability in the tropical Atlantic Ocean. Clim. Dyn.
23, 133–-152 (2004).
18. Giannini, A., Saravanan, R. & Chang, P. The preconditioning of tropical Atlantic
variability in the development of the ENSO teleconnection: implications for the
prediction of Nordeste rainfall. Clim. Dyn. 22, 839–-855 (2004).
19. Chang, P., Saravanan, R. & Ji, L. Tropical Atlantic seasonal predictability: the
roles of El Niño remote influence and thermodynamic air–-sea feedback.
Geophys. Res. Lett. 30, 1501–-1504 (2003).
20. Kiehl, J. T., Hack, J. J., Bonan, G. B., Boville, B. A., Williamson, D. L. & Rasch, P. J.
The National Center for Atmospheric Research Community Climate Model:
CCM3. J. Clim. 11, 1131–-1149 (1998).
21. Zebiak, S. E. & Cane, M. A. A model El Niño–-Southern Oscillation. Mon. Weath.
Rev. 115, 2262–-2278 (1987).
22. Reynolds, R. W. & Smith, T. M. Improved global sea surface temperature
analysis using optimum interpolation. J. Clim. 7, 929–-948 (1994).
Acknowledgements We thank J. C. H. Chiang for a discussion about the
tropospheric temperature mechanism. This work was supported by the NOAA
Climate and Global Change Program and by the NSF Climate Dynamics
Author Information Reprints and permissions information is available at The authors declare no competing financial interests.
Correspondence and requests for materials should be addressed to P.C.
([email protected]).
© 2006 Nature Publishing Group